DailyPapers (@HuggingPapers)
2026-01-15 | โค๏ธ 192 | ๐ 24
NVIDIA Fast-ThinkAct
Efficient VLA reasoning framework that compresses lengthy chain-of-thought into compact latent tokens, achieving 9.3x faster inference while maintaining strong performance on embodied manipulation tasks. https://x.com/HuggingPapers/status/2011656743529353231/photo/1
๐ ์๋ณธ ๋งํฌ
๋ฏธ๋์ด
๐ Related
Auto-generated - needs manual review